Averse

Podcast 2 — In conversation

Speaker A: I've been trying to convince my colleague to adopt a new project management tool and she keeps saying she's averse to change. Is that a fair use of the word?

Speaker B: Perfectly fair. Averse to change is a very natural phrase — it means she has a settled resistance to changing her established ways of working, not just a momentary preference. It implies the feeling is reasonably deep-seated rather than trivial. If she'd just said 'I'd rather not', averse would be too strong. But if she consistently resists new tools and processes, then yes — she's averse to change.

Speaker A: What about 'not averse to'? I've seen that used and it seems like a slightly odd double negative.

Speaker B: It's a very common and deliberate construction. 'Not averse to' means you're actually quite willing or even enthusiastic about something, said with deliberate understatement. 'I'm not averse to an early finish on a Friday' is a polite British way of saying you'd very much welcome it. The double negative softens the admission of enthusiasm — which is quintessentially British. It's a hedge that signals willingness while maintaining composure.

Speaker A: And risk-averse — that's almost become a set phrase in finance and business, hasn't it?

Speaker B: It really has. In finance and economics, risk-averse has a precise technical meaning — a person or investor who, given two options with identical expected value, will prefer the one with less variance or uncertainty. But in everyday business speech it's used more loosely to mean cautious, conservative, or resistant to bold moves. 'Our board is very risk-averse this year' simply means the board is avoiding big bets. The word has migrated from technical economics to general professional vocabulary quite thoroughly.

Speaker A: What's the difference between averse and adverse? I know they're different but I'm never fully confident.

Speaker B: The cleanest way to remember it: averse describes people's feelings — you are averse to something. Adverse describes conditions or events — adverse weather, adverse effects, adverse circumstances. Averse is always about an attitude or inclination; adverse is always about something external that is unfavourable or harmful. So you'd say 'I'm averse to the proposal' but 'the adverse reaction to the medication was unexpected.' They share a Latin root but they've settled into completely different grammatical and semantic roles in English.

Podcast 3 — Prompt engineering

Speaker B: 'Averse' is a surprisingly powerful word in development prompts — particularly for building preference systems, risk assessment tools, user profiling, and recommendation engines where user attitudes and tolerances shape what the system does. Let's go through the scenarios.

Speaker B: First — a risk preference profiler for an investment platform:

Speaker B: PROMPT: Build a risk preference assessment module for an investment platform. Present users with ten scenario-based questions that measure their attitude to financial risk across three dimensions: loss aversion, volatility tolerance, and time horizon. Classify users as highly risk-averse, moderately risk-averse, balanced, or risk-tolerant. Store the classification in a risk_profiles table with user_id, profile_type, scores_json, and assessment_date. Filter investment product recommendations to match the user's risk-averse or risk-tolerant profile. Stack: React, Node.js, and PostgreSQL.

Speaker A: Filtering product recommendations by risk profile — that's the feature that connects the averse-to-risk assessment directly to the user's experience. The three-dimension scoring across loss aversion, volatility tolerance, and time horizon produces a much richer profile than a single risk score.

Speaker B: Good. Now a user preference system with aversion flags:

Speaker B: PROMPT: Add a user preference and aversion system to an e-commerce platform. Allow users to mark product categories, ingredients, materials, or brand attributes as aversions — things they actively want to exclude from recommendations and search results. Store aversions in a user_aversions table with user_id, aversion_type, aversion_value, and created_at. Apply aversion filters to all recommendation engine queries so averse preferences are never surfaced. Include an aversion management UI where users can review and remove their recorded aversions. Stack: React, Node.js, and PostgreSQL.

Speaker A: A dedicated user_aversions table separate from positive preferences — that architecture is clean because it keeps 'what I want' and 'what I'm averse to' in different data structures, which makes the filter logic straightforward and auditable. The aversion management UI gives users agency over their exclusions.

Speaker B: Here's a database schema for a clinical allergy and aversion record:

Speaker B: PROMPT: Design a PostgreSQL schema for a clinical patient management system that records both allergies and treatment aversions separately. The allergies table stores confirmed medical allergies with severity and reaction type. The treatment_aversions table stores patient-reported aversions to medication types, procedures, or care approaches — things the patient is averse to but which are not clinically contraindicated. Both tables link to patient_id. Build an API endpoint that returns a combined aversion and allergy summary per patient for use by clinical staff before any intervention. Stack: Node.js and PostgreSQL.

Speaker A: Separating confirmed medical allergies from patient-reported treatment aversions — that distinction is clinically important. An allergy is a physical contraindication; an aversion is a patient preference. Keeping them in separate tables with a combined summary endpoint means clinical staff see both without confusing the two.

Speaker B: Now a UI design for a risk management dashboard:

Speaker B: PROMPT: Design a risk appetite and aversion dashboard for a corporate treasury management platform. Use a dark navy and amber colour scheme. Display the organisation's current risk-averse thresholds for currency exposure, interest rate risk, and liquidity risk as configurable gauge charts. Highlight any current positions that exceed the averse threshold in amber and those that breach hard limits in red. Add a settings panel where the CFO can adjust the averse threshold values per risk category. Stack: React, Chart.js, and Tailwind CSS.

Speaker A: Configurable averse threshold gauges per risk category — that gives the CFO a live visual representation of how far each position sits relative to the risk-averse boundary the organisation has set. The amber-versus-red distinction between threshold breach and hard limit breach is the critical visual hierarchy for treasury risk monitoring.

Speaker B: One final prompt — a full application:

Speaker B: PROMPT: Build a full-stack dietary preference and aversion management platform for a meal planning service. Users complete an onboarding survey recording dietary requirements, positive preferences, and specific food aversions — ingredients or food categories they are averse to consuming. Store aversions with severity — mild dislike, strong aversion, or medical requirement. The meal planning engine must exclude all averse ingredients from suggested recipes. Weekly meal plans show an aversion compliance score indicating the percentage of meals free from any user aversion. Stack: Next.js, Node.js, and PostgreSQL.

Speaker A: An aversion compliance score per weekly meal plan — that metric gives users confidence that the service is genuinely respecting their aversions rather than occasionally slipping an unwanted ingredient in. The three-tier severity model — mild dislike, strong aversion, medical requirement — means the engine can prioritise hard exclusions over soft preferences intelligently.

Speaker B: The lesson: 'averse' in prompts generates systems that model what users want to avoid, not just what they want. It produces aversion tables, risk-profile classifiers, exclusion filter logic, and compliance scores. The architectural insight is always the same: preferences and aversions are structurally different — one pulls the system towards something, the other pushes it away — and representing them separately produces cleaner, more useful systems.